-
Notifications
You must be signed in to change notification settings - Fork 1
/
notes.txt
27 lines (14 loc) · 699 Bytes
/
notes.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
# Questions about images
- Is the noise level always proportional to total energy?
- Is the detector always hit in a single pixel with some bleed?
Observations for paper:
-The image artifacts appear when using uniform grid for training a model!!!!! (ie is not the fault of
# current problem: no random subpixel placement, makes model learn pixels
TODO:
Compare difference for simulated PSF spread!
-Create superresolution solution (using virtual hits)
optimize data generation (especially for gyroadapted, super slow atm)
generate entire image bach at the same time?
Move trained networks to shared location (where?)
https://keras.io/examples/vision/super_resolution_sub_pixel/
Notes on